Omni-Path beta is live for $Omni-Path holders — Try it out

Local CLI · Zero Slop · Full Verification

The local CLI that forces AI agents to prove their file edits against your actual filesystem.

Zero slop. Full verification.

$Omni-Path
CA:
LFG
omni-path — zsh

Omni-Path can run the agent, load repo rules, or act as the verified file-edit boundary for another agent.

Agent Mode

Use Omni-Path as the agent

Run one task or open chat. For developers who want Omni-Path to call a configured provider and verify any file edits it makes.

$ omni-path init
$ omni-path run "explain this repo"
View docs →
Skills Mode

Teach Omni-Path your repo

Generate and load skills. For teams that want repo rules, review expectations, and risk surfaces available every session.

$ omni-path learn
$ omni-path run --file TASK.md -s repo
View docs →
MCP Mode

Connect another agent

Use Omni-Path over MCP. For MCP clients or external agents that need local dry-runs, verified applies, rollback, receipts, and skill inspection.

$ omni-path mcp
$ swd_dry_run / swd_apply
View docs →

Omni-Path does not ask you to trust a model's story.

It treats each file edit as a claim, compares it with the real filesystem, and records what actually happened.

01

Model proposes actions

Use omni-path chat/run, or bring your own external agent through omni-path mcp or omni-path swd apply.

02

Omni-Path validates the boundary

Input is parsed, paths are constrained to the project, sensitive files are blocked, and high-impact actions require explicit opt-in.

03

SWD applies and verifies

The engine snapshots disk state, applies approved changes, verifies the actual filesystem, and rolls back failed writes.

04

Receipts keep the record

Successful non-dry-run writes produce local receipts with file hashes, provider, MCP, or external-agent metadata, git context, and drift checks.

Not another chat wrapper.

Omni-Path is a local execution boundary for model output, repo rules, receipts, MCP clients, and external-agent file actions.

Strict Write Discipline

Every claimed file edit is checked against the real filesystem with before and after snapshots, hashes, and verification status.

MCP Adapter for SWD

MCP clients can use omni-path mcp as a local tool server for dry-runs, verified applies, receipts, and skill inspection.

One-Shot Run Mode

Use omni-path run for scriptable one-off tasks from inline prompts, files, or stdin while keeping the same SWD, memory, budget, skills, and receipt path.

Repo Skills

Keep project rules in .Omni-Path/skills/<name>/SKILL.md or personal reusable rules in ~/.omni-path/skills. Active skills are recorded in receipts.

Repo Learning

omni-path learn creates a reviewable repo-local SKILL.md from local project signals without running npm scripts, tests, builds, or model calls.

Receipts & Drift Checks

Verified runs write local receipts with file hashes, provider or external-agent metadata, git context, rollback status, and later drift verification.

Install Omni-Path locally.

Then run the command for your flow. Chat and run need a provider key; MCP and external-agent SWD can verify submitted file actions without a Omni-Path model call.

$ npm install -g omni-path
Home / Docs

Omni-Path Documentation

A practical reference for installing Omni-Path, choosing the right entry point, loading repo skills, routing external agents through SWD, and checking local receipts.

Home / Insights

Engineering the Bulletproof AI agent.

Deep-dives into verification protocols, provider orchestration, and the architecture behind zero slop autonomous coding.

Ready to stop AI from lying about your code?

Verified file operations · Multi-provider fallback · Budget tracking · Self-healing memory

Install CLI ⭐ GitHub
Home / SWD

The Omni-Path Engine

A local-first, provider-agnostic CLI runtime. Omni-Path goes beyond simple LLM wrappers by enforcing strict file verification, multi-provider routing, skills, receipts, and MCP/external-agent SWD boundaries.

1

Strict Write Discipline (SWD)

Intent Capture

Captured AI output is treated as a Proposed Intent, not a direct command. Each node must contain explicit metadata: Target Path, Operation, and File Content.

Snapshot Sync

The system performs an instantaneous sync with the Host Filesystem. Actual state overrides internal AI assumptions, creating a verified ground truth baseline.

Correction Turn

If an intent references invalid system state, execution is blocked. Real-time filesystem state is re-injected into the AI context, triggering a self-correction loop.

Atomic Verification

Intents are promoted to COMMIT only when verification yields zero divergence. Ensures atomic, validated writes under controlled conditions.

2

Multi-Provider Orchestration

Fallback-Aware Routing

Omni-Path uses provider health, latency, cost, retry behavior, and fallback settings to choose between configured providers. If the active provider fails, the orchestrator can move to another configured provider.

Adaptive Performance Scoring Exponential Backoff & Circuit Breakers 15s Stream Watchdog Timeout
3

Auto-Healing TDD Loop

Closed-Loop Execution

When the --test-cmd flag is passed, Omni-Path automatically runs your test suite after applying SWD verified file changes.

Error-Driven Fixes

If tests fail, the stdout/stderr is captured, truncated safely, and fed back into the agent context for an automatic correction turn.

Anti-Thrashing Guards

Safety mechanisms detect identical test outputs across retries or regression in failure counts to prevent runaway token consumption.

Budget Enforced

Every test healing loop respects the global SessionBudget, gracefully saving state if token or turn limits are approached.

4

YAML Skills Protocol

Inject targeted operational context dynamically using local SKILL.md files. Skills shape repo behavior, review expectations, provider preferences, and budget settings for the active session.

5

MCP and External-Agent SWD Boundary

Local MCP Tool Server

MCP clients can run omni-path mcp and call SWD dry-run/apply, receipts, and skill inspection tools over local stdio.

Automation Output

--json returns machine-readable status, approved and rejected action counts, verification results, and receipt metadata for scripts and CI-style tooling.

Security Defaults

Sensitive paths are blocked by default. Deletes and command-surface files require explicit opt-in.

Receipts

Successful non-dry-run applies write local SWD receipts with external agent metadata, file hashes, rollback status, and drift verification data.

Upgrade Your Workflow

Deploy the Omni-Path in your development environment to transform generative AI into a reliable, verifiable, and cost-efficient engineering partner.

GitHub